CCI vs Williams %R

Michael Whitman · Trading since 2017, full-time since 2020 · Reviewed 2026-09-08

Two oscillators that both break the conventions people expect. One has no boundaries; the other runs backwards. Understanding which is which is most of what this comparison is for.

What each one is

The commodity channel index measures distance from a moving average, divided by typical deviation, so its scale adjusts to how much the instrument normally moves. The commodity channel index covers it.

Williams percent R measures position within the recent high-low range, counting down from zero at the top to minus one hundred at the bottom, unsmoothed. Williams percent R covers it.

One adapts and the other does not. The normalisation in the first is a genuine feature; the second’s lookback is a fixed window with nothing scaling it.

Where they differ

Whether the reading adapts. Dividing by typical deviation means a busier market does not automatically produce larger readings. A raw range position has no such adjustment.

Whether there is a ceiling. One can print any value; the other cannot leave its range. That changes what happens in a strong move — one keeps rising, the other pins.

How much smoothing there is. Some in the first; none at all in the second, which makes Williams percent R the noisiest reading in this family.

Which way the scale runs. Up from weakness in one, down from strength in the other. Copying a threshold rule between them produces exactly the wrong trade.

Where they agree

Both are computed from the same price series. Neither adds information from outside it, so they cannot function as independent checks on each other.

Both lag. Every value came from bars that have already closed, and neither can turn before price does.

Both fail in a range. On this site’s shared series direction runs average 2.01 bars with a longest of 11, and short runs cross any threshold repeatedly on either.

And both are misread more often than most. One because its levels are not limits, the other because its scale is upside down — and neither problem is a fault of the market.

Which one to use

Run the commodity channel index when you want the reading to adapt. Normalising by typical deviation means the tool stays comparable to itself when a market changes character.

Run Williams percent R when you want the fastest raw range reading. No smoothing means no added lag, and every turn arrives as early as it possibly can.

Run the stochastic instead if you want that same measurement steadied. It is the smoothed, conventionally scaled version of what Williams percent R reports, and far better supported.

And run one, not both. They read the same bars, so a chart carrying both has two unconventional scales describing one thing.

Why both are misread so often

Because the conventions do not apply. People carry habits from bounded, upward-counting oscillators, and neither of these tools is that.

And because the charts look familiar anyway. Both produce a line in a panel that resembles every other oscillator, so nothing on screen warns you the axis means something different.

What to check on each

On the normalised one, look up what your instrument usually prints. Without that range the conventional thresholds are numbers from somewhere else, and they may be crossed weekly.

On the raw one, confirm which end is strength. Near zero means the close is at the top of its range, which is the opposite of the habit almost every other oscillator builds.

On both, count the signals over a full year. On this site’s shared series direction runs average 2.01 bars, so any threshold rule fires far more often than the market changes direction.

And on both, leave the length alone once set. Changing it after a losing run reshapes every past signal and leaves you with no accumulated evidence about either configuration.

The original data

Of the 24,971 unique videos in research/search-study-corpus.jsonl, no title compares these two directly — this pair is constructed from two subjects the corpus covers separately. Separately, the commodity channel index appears in 448 titles at a median of 9,318 across 344 channels, and Williams percent R in 23 at a median of 6,222 across 23. The counts come from site/corpus_count.py.

448 videos across 344 channels against 23 across 23. Twenty times the coverage for the normalised tool, and the other averages exactly one video per channel — nobody makes a second one, which is the clearest signal available about how widely it is actually used.

The answer to the question on that chart is that both are saying the same thing. Extending on one and near zero on the other are both descriptions of strength — and if that reads as disagreement, the inverted scale has done what it usually does.

When it fails

The failure is reading the inverted scale the wrong way round, and every trade comes out backwards. Near zero means the close sits at the top of its recent range, which is strength. Carried over from any conventional oscillator, that number reads as weakness, so longs are taken at lows and shorts at highs with complete confidence. The chart offers no warning because the line looks exactly like every other oscillator line, and the mistake survives until somebody checks the definition.

The second failure is treating a conventional level as a limit. The other scale is open.

A third is running both for confirmation. They read the same bars.

A fourth is expecting to smooth the raw one. There is nothing to smooth.

A fifth is carrying thresholds between instruments. Neither scale transfers cleanly.

And a sixth is adopting a tool nobody discusses. You will debug it alone.

Related

The commodity channel index covers the normalised distance reading. Williams percent R covers the raw inverted range reading. And stochastic is the smoothed, conventionally scaled version of the second.

This page is educational, not financial advice. Test every idea on your own charts before risking money.